2490

ACTIVE FAULTS IN DAM FOUNDATIONS: AN UPDATE

Clarence R ALLEN1 And Lloyd S CLUFF2

SUMMARY

This paper updates the 1974 Geotechnique paper on the same subject by Sherard, Cluff, and Allen.
Geologists and seismologists are now able to recognize the presence of, and assess the degree of
activity of, faults in dam foundations far more effectively than only a few years ago, thanks to a
variety of new neotectonic tools. Similarly, engineers are increasingly able to accept possible
foundation displacements with a variety of innovative measures in dam design. Brief case studies
of faults beneath dams are presented, and some mitigative measures are described, for Auburn
Dam site (USA), Clyde Dam (New Zealand), Eastside Reservoir USA), Lauro Dam (USA),
Matahina Dam (New Zealand), Ridgway Dam (USA), Seven Oaks Dam (USA), Steno Dam site
(Greece), and Tarbela Dam (Pakistan).

INTRODUCTION

The purpose of this paper is to bring up to date the 1974 paper by James Sherard and the two present authors,
"Potentially active faults in dam foundations" [Sherard et al., 1974]. The present paper is dedicated to Sherard,
who was a highly valued friend and colleague. This paper concentrates primarily on recent developments in
methods of fault identification and activity assessment, rather than mitigative measures in engineering dam
design.

To the authors' knowledge, there has as yet been no historic case of an operating dam being displaced by a fault
during an earthquake, although there have been some "near misses." This good record of worldwide performance
is particularly remarkable in view of the fact that manyif not mostdams are located in river canyons whose
courses are controlled by preferential erosion along underlying faults and joints.

WHAT IS AN ACTIVE FAULT?

Not surprisingly, almost all foundations for large dams display some faults, however minor, and the geologic and
seismologic challenge is to determine whether such faults are likely to rupture during the life of the structure
(i.e., are they "active"?) and, if so, with what displacements, with what geometries, with what magnitudes, and
with what likelihoods. Over the years, many authors have discussed various definitions of "active" and "inactive"
faults. Suffice it is to say that modern geologic studies, together with vastly improved age-dating capabilities,
have demonstrated unequivocally that there are all degrees of fault activity, and any categorization into active
and inactive features is necessarily arbitrary. In worldwide dam-design practice, repeat times on faults of
significant earthquakes of a few thousand years, or a few tens of thousands of years, are often used to distinguish
between faults one wishes to worry about and those of no concern.

HAZARD ASSESSMENT OF FAULTS

The last few years have been marked by numerous significant improvements in fault recognition and hazard
assessment techniques, particularly in improved methods of absolute age dating that have made possible
meaningful probabilistic assessments that were formerly impossible. These techniques have been discussed in
detail elsewhere [e.g., Yeats et al., 1997] and are not repeated in detail herein, nor is a discussion of the merits of
probabilistic vs. deterministic methodologies.

1
Seismological Laboratory, California Institute of Technology, Pasadena, Calif. 91125, USA E-Mail: [email protected]
2
Geosciences Department, Pacific Gas and Electric Company, San Francisco, California 94177, USA E-mail: [email protected]
Perhaps the most critical parameter of value in determining the degree of activity of a fault is its slipe rate, that
is, its average rate of slip (usually expressed in mm/yr) over a sufficiently long time period to include several
large earthquakes. This, of course, necessitates age measurements of some displaced feature or features, such as
strata, stream courses, river terraces, erosion surfaces, etc. Age-dating of young rocks, which is essential to
determining degree of fault activity, formerly depended almost solely on recovering organic materials which
could be dated by radiocarbon methods. But a myriad of new techniques is now available and others are under
vigorous development. Particularly important are novel methodologies for dating the length of time the ground
surface has been exposed (e.g., a stream terrace surface), using cosmogenic nuclides. There has always been a
hope that, instead of dating the rocks or surfaces offset by a fault, something within the sheared, clayey material
making up the fault itselfthe so-called fault gougemight be able to reveal the time of the latest or earlier
movements. Novel methods are being investigated and attempted, but no technique utilizing the gouge itself has
as yet been proven generally applicable. A summary of available techniques and their limitation is given by
Yeats et al. [1997, ch. 6].

THE PROBLEM OF BRANCH FAULTS

The problem of branch faults—those that are branches of a more significant "main" feature—has long plagued
geologists, seismologists, engineers, and regulators. If the main fault has a sufficiently low degree of activity to
dismiss it as a hazard, and if there is no indication that a branch is more active, then it is usually straightforward
to put the branch fault in the same non-hazardous category as the main fault. But a more troublesome situation is
that in which the main fault has a high degree of activity. Should all branch faults then be considered equally as
hazardous as the main break? If so, the quandary immediately arises as to what features really are to be
considered branches, and what kinds of displacements are to be assumed relative to those on the main fault?

In the authors' opinion, no matter how high the degree of activity of the main fault, the possibility of future
displacements on a branch fault can be dismissed if the branch fault can be demonstrated to have an acceptably
low degree of activityfor example, if it is covered and truncated by unbroken late Quaternary strata of
hundreds of thousands of year age. That is, if the branch fault has not broken sympathetically with the main fault
over a sufficiently long period of time, there is no reason to assume that it will do so in the near future. More
difficult, however, is the not-unusual situation in which a branch fault may show no evidence of activity one way
or the otherthat is, there are no young strata in the local area to demonstrate whether or not they are would be
broken if present. This is a situation where geologic judgment necessarily comes into play. A principal line of
evidence is the geometrical relationship of the branch fault to the main fault. How close are they in attitude? Do
they actually join? Could the same stress field actuate both? How conservative does one wish to be? This was the
situation faced at Auburn Dam, USA, Clyde Dam, New Zealand and Steno Dam site, Greece (see below).

FAULT CREEP VS. ABRUPT FAULT DISPLACEMENT

Fault creep is gradual, continuing displacement, usually at a very low rate. This behavior is in contrast to the
abrupt fault displacement causing an earthquake, which often takes place within only a very few seconds,
associated with displacements of several meters and particle velocities of as much as 1 m/sec [Wallace, 1984;
Heaton, 1990]. Fault creep was first recognized in 1956 on a segment of the San Andreas fault in California,
USA. Subsequent careful investigation of active faults throughout the world, however, indicates that creep is a
relatively rare phenomenon, even along most of the San Andreas fault. Only on one limited segment of the North
Anatolian fault in Turkey, and possibly on a segment of the Xianshuihe fault in China, has continuing creep been
documented, in addition to reports of two faults that cut the Bajina Basta Dam and the Lipovica Dam in
Yugoslavia [Bozovic and Markovic, 1999]. In the authors' opinion, continuing fault creep is a sufficiently rare
phenomenon so that it can usually been dismissed in hazard studies unless direct evidence to the contrary exists.

SEISMOLOGICAL EVIDENCE OF FAULT ACTIVITY

Another difficult problem for the geologist and seismologist is the degree to which locations of instrumentally
recorded earthquakes should play a role in deciding whether rupture on a fault is credible. In actuality, not all
faults on which major earthquakes have occurred historically (e.g., the San Andreas fault in California) are
characterized by continuing seismic or microseismic activity along the same fault segment. And some major
earthquakes have occurred on faults with very little, if any, precursory seismicity or microseismicity within the
previous few years. Therefore, the presence or absence of seismic activity must be viewed with great caution in
determining the degree of activity of a given fault, [Allen et al., 1965; Yeats et al., 1997]. Likewise, one should
be very careful in postulating fault locations from hypocentral locations. As applied to dam foundations, it
should be emphasized that, inasmuch as surface fault rupture is the hazard of principal concern, the demonstrated

2 2490
absence of surficial late Quaternary displacement on a fault is usually an adequate guarantee of similar behavior
in the near future regardless of whether or not micro-earthquake epicenters are present at depth beneath the
structure.

FAULTING ASSOCIATED WITH RESERVOIR-TRIGGERED EARTHQUAKES

Although there have been probably hundreds of cases worldwide of reservoir-triggered earthquakes, mostly of
very small magnitudes, the authors are aware of only two instances where the triggered events are known to have
been associated with unequivocal surface faulting. One of these was at Koyna, India, in association with the M =
6.3 triggered event of 1967 [Cluff, 1977], and the other at Oroville, California, USA, in association with the M =
5.7 event of 1975 [Clark et al.¸1976]. In both cases, rupture was along a pre-existing fault with earlier
displacements of Holocene age, although in neither case did the fault actually pass beneath the dam.

It was suggested by Allen [1982] that reservoir-triggered earthquakes possibly could occur on a fault otherwise
considered to be "inactive," because "the reservoir may produce a stress distribution quite unlike that which the
area has experienced for many thousands of years." While perhaps theoretically possible, it must be recognized
that, to date, only two out of literally hundreds of worldwide reservoir-triggered earthquake sequences have
been associated with unequivocal surface ruptures. In the current opinion of the authors, the likelihood of a
reservoir-triggered earthquake with significant surface displacement on a fault with proven absence of late
Quaternary displacements is exceedingly remote, particularly if the associated reservoir is not deep and large.

RESPONSE TO DISCOVERY OF FOUNDATION FAULTS

When a fault of suspected displacement hazard is discovered in a dam foundation, either during initial
exploration and excavation, or during reassessment at some later date, responses seem to fall into several
categories: (1) If still under planning, the design of the dam may be altered and/or the dam's alignment adjusted
to minimize the risk. Such was the case at Cedar Springs Dam, California, USA [Sherard et al., 1974] and at
Clyde Dam, New Zealand (see below). (2) Through geologic studies involving age-dating of affected foundation
rocks, it may be possible to determine that the slip rate on the fault is sufficiently low, and/or the time of last
movement is sufficiently great, so that no significant danger to the dam is present. That is, for all practical
purposes, the fault is determined to be "dead." Such was the case at the Eastside Reservoir Project, California,
USA (see below). (3) Engineering studies may indicate that the dam can safely withstand the largest expected
fault displacement. Such was the case at Tarbela Dam (Pakistan), and at Seven Oaks Dam, California, USA (see
below). (4) In an existing dam, modification and strengthening may be possible to reduce to risk to a low level.
Such as was the case at Matahina Dam, New Zealand (see below). (5) The site may be abandoned, at least for the
type of dam being planned, either before or after construction starts. Such was the case at Auburn Dam,
California, USA (see below).

SELECTED CASE STUDIES

The following brief case studies represent a few that have come to the attention of the authors. There are
undoubtedly many others examples worldwide that could also have been mentioned, and some of these are
summarized by Bozovic and Markovic [1999]. Leps [1989] not only discusses other examples, but also presents
a good discussion of a variety of defensive design measures. Dams that were specifically discussed by Sherard et
al. [1974] are not discussed again herein.

Auburn Dam site, California, USA: Construction was started in 1967 by the U.S. Bureau of Reclamation
(USBR) on what was planned to be the world's longest double-curvature thin-arch concrete dam, a 210-m-high
structure across the American River 45 km northeast of Sacramento, California. Although it was recognized at
the time that the predominantly pre-Cretaceous metamorphic rocks of the foundation were highly deformed and
faulted, it was not considered by the USBR that any of these faults were "active," and the region's historic and
instrumental seismicity was relatively low. Geologic mapping of the excavated foundation was carried out
competently and in great detail, and none of the many faults therein showed evidence of youthful displacements.
However, none of these faults was mantled in the dam footprint area by younger rocks that could have revealed
youthful displacements if they had, in fact, occurred. In the meantime, the 1975 Oroville earthquake (M = 5.7)
occurred along a member of the same fault system about 50 km northwest of Auburn, with surface displacements
of as much as 5 cm [Clark et al., 1976], albeit perhaps triggered by the nearby Oroville Reservoir [Toppozada
and Morrison, 1982]. Furthermore, contemporary regional studies along the Foothills fault system, which
included the Auburn area, had revealed unequivocal evidence of late Tertiary displacements at several localities,
as well as some indications of late Quaternary displacements [Woodward-Clyde Consultants, 1977]. These

3 2490
studies stimulated subsequent neotectonic investigation of faults close to the Auburn area itself, which revealed
several instances of surface fault ruptures of less than 100,000-year age, including one at a trench locality only
800 m from the dam excavation. Literally dozens of trenches were excavated across these faults. Although there
was considerable controversy about the interpretation of the exposures in these trenches and their relevance to
the faults within the dam footprint, a number of outside consultants (including, independently, both authors of
this paper) agreed that fault movements beneath the dam were possible during a maximum earthquake, with
displacement estimates ranging from 2 cm to about 1 m. As a result, and also taking into account reservations
expressed by the U.S. Geological Survey and the California Division of Mines and Geology, the Auburn Dam
project was cancelled after many hundreds of millions of dollars had already been expended on major foundation
excavation and concrete dental work. There was simply not sufficient confidence that a concrete thin-arch dam
could safely withstand even small foundation displacements.

Clyde Dam, New Zealand: Clyde Dam is a 102-m-high concrete gravity dam across the Clutha River, New
Zealand, completed in the early 1990's. The nearby Dunstan fault had long been recognized as extending at least
60 km north from the dam area, with Holocene-active segments within 3 km of the dam itself [Beanland, et al.,
1986]. The Dunstan fault was the principal feature guiding the design of the dam for seismic shaking, with a M =
7-7.5 earthquake assumed to occur on it with an average frequency of about once in 8,000 years. The main fault
did not extend directly into the dam foundation, but several apparent branches were mapped nearby. During
preliminary exploration for the dam, but before construction started, the River Channel fault was identified in the
deepest part of the foundation, and, although its trend was roughly perpendicular to that of the Dunstan fault, and
it could be followed for only some 780 m, it was deemed credible that it could move in sympathy with a major
earthquake on the Dunstan fault. The reasoning was apparently related to its proximity to the Dunstan fault and
to its wide shear zone within the foundation [Hatton et al., 1991]. Furthermore, the Dunstan fault is a thrust fault
with complicated near-surface fracturing, and elsewhere along its trace are many minor faults within 2-3 km that
have demonstrable late Quaternary displacements [Beanland et al., 1986]. Others pointed out that the strain field
near the termination of a major active fault may be very complicated, so the perpendicular orientation of the
River Channel fault might not in itself rule out possible sympathetic movement [Nelson, 1984]. The New
Zealand Geological Survey estimated that up to 20 cm of slip could occur on the River Channel fault in
sympathy with a major earthquake on the Dunstan fault. The design of the dam was therefore modified by
including an appropriately located slip joint directly above the fault, with a rubber-sealed, steel-sheathed wedge
plug, 100-m high, to accommodate as much as 2 m of strike slip and 1 m of dip slip movement on the River
Channel fault [Hatton et al., 1987; 1991]. The intent was "to keep the joint water tight during normal operation
and to limit the flow in the event of fault displacement" [Hatton et al., 1987]. The powerhouse, sluice intake,
and spillway were also relocated to avoid the fault and slip joint. .

Eastside Reservoir Project, California, USA: Currently nearing completion, the Metropolitan Water District's
Eastside Reservoir Project, near Hemet, California, is the largest ongoing dam construction project in the U.S. It
involves three embankment dams enclosing a formerly dry valley. The dams are founded largely on
metamorphic and igneous rocks of more than 100-million-year-old age. The Project is located at a closest
distance of 8 km from the highly active San Jacinto fault—a major member of the San Andreas fault system—
but no Holocene faults are known to be present in the reservoir area itself. The largest dam (West Dam) is some
3300 m long and 87 m high, but its construction involved excavation below grade as much as 25 m to
unweathered bedrock, except for 3 relatively narrow and still deeper ancient channels filled with highly
consolidated Quaternary alluvium. Several relatively minor faults were exposed, not unexpectedly, in the ancient
metamorphic bedrock of the foundation, but one of these was found to extend into the adjacent channel-filling
sediments, with a maximum vertical separation of about 5 cm in the old alluvial deposits [Metropolitan Water
District, 1996]. In a series of 11 fault-crossing trenches, the fault was followed in the old alluvium—sub-parallel
to the ancient channelfor about 320 m, at which point the fault was truncated by an overlying unbroken
younger alluvial unit. Fortunately, within the younger alluvial unit several beds were capped by ancient soils
whose ages could be estimated on the basis of diagnostic pedogenic features. Based on these features, it was
possible to infer that the latest displacement on the fault took place more than 200,000 years ago. A number of
vertebrate fossils in the sequence also supported a relatively great age for the old alluvium. On the basis of the
age of the most recent faulting event, as well as the relatively small displacements involved, it was judged that
the fault created no concern for the safety of the Project.

Lauro Dam, California, USA: Lauro Dam, part of the USBR Cachuma Project, is a 42-m-high dam within the
city of Santa Barbara, California. When the foundation was excavated in 1951, a fault was recognized within the
dam footprint that juxtaposed alluvial fan deposits against Miocene sedimentary rocks. Only limited geologic
work was carried out at that time, but in 1998 an extensive exploration program was initiated, including 2
trenches and more thorough geologic mapping, in order to determine the degree of fault activity and its possible

4 2490
displacement capability [Anderson, 1999]. Although the trenches and excavations at the site reaffirmed the
presence of the fault, analyses of regional geologic data studies were necessary to determine the critical hazard-
assessment parameters. The foundation fault appears to be part of the 50-km-long Mission Ridge-Arroyo Parida-
Santa Ana fault system, with predominantly left-lateral strike-slip displacement. Uncertainties are high with
regard to slip rate and slip-per-event, although the length of the fault suggests significant displacements are
possible. Anderson [1999] concluded that " . . . , the maximum surface displacement in the foundation of Lauro
Dam could be as high as 2 m. Assuming a slip rate of 1 mm/yr, the annual probability of 1 m or more of surface
rupture through the embankment and outlet works of Lauro Dam could be higher than 0.002/yr." The need, if
any, for possible mitigative measures has not yet been decided upon.

Matahina Dam, New Zealand: Faulting at Matahina Dam, an 80-m-high embankment structure on the
Rangitaiki River of the North Island of New Zealand is described in detail by Mejia et al. [1999] and Freeman et
al. [this volume]. When the dam was constructed in the 1960's, several prominent faults parallel to the river trend
were observed and mapped in the mid-Quaternary sedimentary foundation rocks exposed in the core trench.
Although carefully mapped, none of these faults was observed to displace the overlying valley-fill alluvium. It
was only later recognized—after completion of regional geologic studies elsewhere in New Zealand—that these
faults comprised the northernmost splays of the Alpine fault system, one of the world's longest and most active
regional strike-slip faults. This realization, together with the fact that the dam was slightly damaged by the very
nearby Edgecumbe earthquake of 1987 (M = 6.3), albeit on a different fault system, led to a re-assessment of the
dam's earthquake safety. An intensive geologic study of the neotectonics of the fault system was undertaken
within some 40 km of the dam, which included a number of trench excavations and was aided by the fact that
several Holocene volcanic ash deposits in the damsite area could be confidently identified and dated. These
studies led to the conclusion that at least some of the fault splays underlying Matahina Dam were highly active
and could be expected to suffer oblique displacements of up to 3 m with an estimated exceedence probability of
between 1/6000 and 1/11000 annually [Woodward-Clyde Consultants, 1996]. Engineers judged that the dam
could not safely withstand a displacement of this magnitude, partly due to the somewhat brittle nature of the
existing core. It was deemed impractical either to remove the dam or completely to replace it, so an innovative
mitigative solution was devised in which the downstream shoulder of the dam was excavated, a leakage-resistant
filter and transition zone constructed within the dam, and a thick rock buttress was placed on the downstream
face of the dam to guarantee that, even if the integrity of the core was compromised during a future fault
displacement—however unlikely—flow would be impeded to the extent that no excessive downstream flooding
would result [Mejia et al., 1999]. As part of the strengthening program, a cutoff trench for the new filter zone
was excavated some 100 m downstream of the original core trench. As expected, several faults were observed in
the underlying bedrock in this new trench, some of which undoubtedly corresponded to faults mapped in the
original core trench nearby. Three of these faults, however, were now observed to displace the overlying basal
valley-fill alluvium, with displacements as great as 1 m. The latest movement on one of the faults could be dated
as between 3,690 and 10,600 years, B.P. [Woodward-Clyde Consultants, 1998; Freeman et al., this volume].
This documentation of very youthful faulting was, in a sense, satisfying, in that it confirmed the wisdom of
having concluded that the faults had a high degree of activity and truly represented a risk to the integrity of the
former dam.

Ridgway Dam, Colorado, USA: Ridgway Dam, a USBR embankment structure across the Uncompahgre River
in southwestern Colorado, was completed in 1987. Some debate still exists, however, concerning the probability
of surface faulting through the foundation, and it represents an interesting case of the relative importance of
geologic vs. seismologic data. During excavation, the Cow Creek fault was discovered in the right abutment and
outlet works area, but trench exposures revealed that the fault did not break overlying glacial outwash of middle
Pleistocene age (>130 ka)[Ake et al., 1997]. Nevertheless, the Cow Creek fault, as well as others in the vicinity,
is debatably associated with minor seismicity that largely post-dates reservoir impoundment. An interesting
probabilistic analysis was carried out, based on a logic-tree analysis and the recorded seismicity, which
tentatively suggests that surface displacements of several tens of centimeters might occur with recurrence
intervals on the order of a thousand years. Many assumptions are necessarily involved, including an exponential
recurrence curve truncated at magnitude 6.3 (i.e., no “characteristic” earthquake), and use of the Wells and
Coppersmith [1994] regression relationships, with uncertainties. Any displacement during the project life is,
needless to say, difficult to rationalize with the local geologic evidence of no surface displacement in more than
130,000 years. In this case, it is argued that, since the current seismicity commenced with the initial reservoir
filling, the late Quaternary geologic history of the Cow Creek fault may be irrelevant to the reservoir-triggered
events [Jon Ake, personal communication]. Studies are continuing.

Seven Oaks Dam, California, USA: Seven Oaks Dam is a U.S. Army Corps of Engineers (USCE) project,
envisaged initially primarily for flood control, that is currently nearing completion in the canyon of the Santa

5 2490
Ana River 18 km east of San Bernardino, California. The curved embankment structure will be 168 m high. It
was fully recognized from the outset that the damsite was located between two major strands of the highly active
San Andreas fault—the South Branch (the principal and most recently active strand) 2 km to the south, and the
North Branch less than 1 km to the north. An earthquake of about magnitude 8 on the South Branch was
assumed to occur during the life the structure, and although the North Branch is also "active," its slip rate was
found to be 1 or 2 orders of magnitude less than that of the South Branch. Since the damsite is virtually within
the San Andreas fault zone, the question of active faults in the foundation arose early in the planning stage. One
the authors (CRA) recommended to the USCE in 1984 that the dam be designed "on the assumption that as much
as 4 feet (1.2 m) of surficial displacement in any direction could take arbitrarily beneath the facility," as
cumulative slip on one or more faults. The dam and its appurtenant structures are so designed, with a wide core,
transition zones, and drains. When excavation commenced, it quickly became clear that several significant faults
did, in fact, cut the foundation, including one fault beneath and parallel to the core that showed evidence of
displacement of the valley-fill alluvium. A detailed field study [Sadler and Rasmussen, 1991] confirmed that
three sequential valley-fill units were cut by the fault, with the amount of offset (a few centimeters) decreasing
up-section. Although no datable materials were obtained, several features suggested that the fault may have
moved within the past 35,000 years. Sadler and Rasmussen [1991] recommended that the fault "should be
considered capable of moving in the near future," and estimated that slip during a single event "may approach 20
cm of oblique right-slip movement." The length of the fault as a continuous feature is now known to be at least
1.4 km, and exploration still continues in the attempt to determine its total length. Seven Oaks Dam is an
example of a situation where both intense shaking and foundation faulting are recognized as being very
credible—if not likely—during the life of the project. Nevertheless, in this light, engineers are satisfied that the
dam and appurtenant structures have been designed and constructed with adequate seismic safety margins.

Steno Dam site, Greece: Steno Dam was a planned 185-m-high double-curvature thin-arch dam to be built
across the gorge of the Arakhthos River in northwestern Greece [Gilg et al., 1987; Pedro and Azenedo, 1988],
although the dam has not yet been built, apparently for economic reasons. A conspicuous fault—the Steno
faultunderlay the deepest part of the gorge, and alternative sites or dam types were not deemed practical. The
Steno fault is not itself a major feature and did not display evidence indicating late Quaternary displacements.
However, other "main" mapped faults nearby were thought to have a reasonably high degree of activity. At least
one, shown as an "active" thrust fault by King et al. [1993], is some 30 km in length, and many other similar
active faults characterize northwestern Greece. A microearthquake study of the dam region located many events,
but Kiratzi et al. [1987] concluded that the activity "indicates deformation over a broad volume of the crust
rather than along well-defined faults.” Although the Steno fault was thought to be "a secondary fault" to the main
nearby faults, it was judged on the basis of the regional neotectonic history and historical seismicity of
earthquakes up to magnitude 6.5 that "displacements on the Steno fault cannot be excluded, but they would be
limited to some decimeters only." [Gilg et al., 1987]. The dam was designed with such a displacement in mind,
mainly by introducing a complicated system of grooved slip joints to create adequate "fexibility" to the structure
to withstand the estimated foundation displacements. Subsequently, the entire dam and abutments were
successfully tested under simulated earthquake displacement loads in a sophisticated model constructed at the
Laboratorio Nacional de Engenharia in Portugal. Details of the proposed design and the model testing are given
by Gilg et al. [1987] and Pedro and Azevedo [1988].

Tarbela Dam, Pakistan: During the initial exploration for Tarbela Dam—the world's largest embankment
structure—it was discovered that the valley alluvium of the Indus River, upon which the dam rests, was
considerably thicker on one side of the 2-km-wide valley than on the other. This variation was associated with a
steep buried bedrock escarpment, and in the initial planning, it was argued that this escarpment might represent
simply a buried steep bedrock wall of the highly incised glacial Indus River, or perhaps a featured scoured by
glacial ice itself . Upon detailed drilling, however, it turned out that the 200-m-high escarpment was actually
overhanging [see Yeats and Hussein, 1989, Fig. 7], and the rock properties were such that it could not have
existed as a freestanding slope. Thus it was argued that the buried escarpment must represent, at least in part, a
fault contact. And since the abutting gravels are post-glacial in age, the fault would have to have a relatively high
degree of activity. Subsequently, it was mapped as the Darband fault [Calkins et al., 1975], and its mapped
length and segmentation length were used to suggest a maximum earthquake on it of M = 6.5, using regression
techniques similar to those later summarized by Wells and Coppersmith [1994]. Similar regressions were also
used to suggest a maximum displacement on the fault of about 1 m during such an event. Engineers judged that
the nature and thickness of the dam's core, transition zones, and drains would allow the structure safely to
withstand such a displacement.

6 2490
 CONCLUSIONS

• There is no standard definition of what constitutes an "active" fault, which reflects the geologic reality that
all degrees of fault activity do, in fact, exist. Any distinction between active and inactive faults is necessarily
arbitrary.
• In worldwide practice, faults in dam foundations that are estimated to have ruptured on the average of once
in every few thousand, or every few tens of thousands of years, have typically been considered worthy of
engineering consideration.
• At least some engineers appear to be gaining confidence that, with innovative new techniques, dams of
virtually all types can be designed to accommodate moderate foundation fault displacements without major
failure.
• Owing to advances in the ability to obtain absolute ages on geologic materials, mainly through new
geochemical techniques, degrees of fault activities can now be estimated far better than only a few years ago. A
critical parameter in both probabilistic and deterministic assessments is long-term fault slip rate.
• Field studies of faults in dam foundations demand the expertise of geologists trained and experienced in
neotectonic studies. Classical geological maps, however competently done, often have only limited relevance to
earthquake hazard assessments.
• Seldom are adequate field exposures available in the dam footprint area itself to characterize fully the
neotectonic environment of a proposed dam. It is virtually always necessary to carry out such studies, including
trenching of suspect faults, throughout a much wider area than that of the dam footprint—typically extending
over many tens of kilometers.
• Seismologic studies of earthquakes in dam areas can play an important role in safety evaluations,
particularly in establishing seismic shaking parameters such as maximum ground accelerations and velocities to
be expected. But field geologic efforts are nevertheless critical in estimating the likelihood of surface fault
rupture through a dam foundation. The abundance or absence of microearthquakes may have little relevance to
the probability of large local earthquake associated with surface fault rupture.

ACKNOWLEDGMENTS

The authors appreciate the helpful comments and permissions to publish by J. Ake, L. Anderson, A. Areta, K.
Cato, P.F. Foster, M. Gillon, J. Lowe, III, and D. Lukesh.

REFERENCES

Ake, J., Vetter U., and Hawkins, F. (1997), Seismicity and Quaternary faulting near Ridgway, Colorado,,
Colorado Geol. Survey Geohazards Conf. Guidebook.
Allen, C.R. (1982), "Reservoir-induced earthquakes and engineering policy", Calif. Geology, 35, pp 248-250.
Allen, C.R., St. Amand, P., Richter, C.F., and Nordquist, J.M. (1965), "Relationship between seismicity and
geologic structure in the southern California region", Bull. Seismol. Soc. Am., 55, pp 753-797.
Anderson, L.W. (1999), Interim assessment of surface faulting potential at Lauro Dam, Cachuma Project,
California, U.S. Bur. Reclamation, Tech. Service Center, Tech. Memorandum D8330-99-011.
Beanland, S., Berryman, K.R., Hull, A.G., and Wood, P.R. (1986), "Late Quaternary deformation at the Dunstan
fault, central Otago, New Zealand", Roy. Soc. New Zealand Bull, 24, pp 293-306.
Bozovic, A. and Markovic, M. (1999), Neotectonics and dams – Guidelines, ICOLD Committee on seismic
aspects of dam design.
Calkins, J.A., Offield, T.W., Abdullah, S.K.M., and Ali, S.T. (1975), "Geology of the southern Himalaya in
Hazara, Pakistan, and adjacent areas", U.S. Geol. Survey Prof. Paper 716-C, pp 1-29.
Clark, M.M., Sharp, R.P., Castle, R.O., and Harsh, P.W. (1976), "Surface faulting near Lake Oroville,
California, in August, 1975", Bull. Seismol. Soc. Am., 66, pp 1101-1110.
Cluff, L.S. (1977), Notes of a visit to Koyna Dam, India, January 14, 15, 16, 1977, Unpubl. rept. to U.S. Bureau
of Reclamation.

7 2490
Freeman, T., Gillon, M., Berryman, K., Moriwaki, Y., and Somerville, P. (this volume), Matahina Dam –
Surface fault rupture design criteria.
Gilg, B., Indermaur, W., Mattley, F., Pedro, O., Cruz Azevedo, M., and Ferreira F. (1987), Special design of the
Steno arch dam in Greece in relation with possible fault movements, Proc. Internatl. Symp. on earthquakes and
dams, Beijing, 1, pp 202-218.
Hatton, J.W., Black, J.C., and Foster, P.F. (1987), "New Zealand's Clyde power station", Internatl. Water Power
and Dam Construction, 39, pp 15-20.
Hatton, J.W., Foster, P.F., and Thomson, R. (1991), "The influence of foundation conditions on the design of
Clyde Dam", 17th Internatl. Cong. On Large Dams (Vienna), Q.66-R.10, pp157-178.
Heaton, T.H., (1990), "Evidence for and implications of self-healing pulses of slip in earthquake rupture", Phys.
Earth Plan. Interiors, 64, pp 1-20.
King, G., Sturdy, D., and Whitney, J. (1993), "The landscape geometry and active tectonics of northwest
Greece", Geol. Soc. Am. Bull., 105, pp 137-161.
Kiratzi, A.A., Papadimitrou, E.E., and Papazachos, B.C. (1987), "A microearthquake survey inthe Steno Dam
site in northwestern Greece", Annal. Geophys. B, 5, pp 161-166.
Leps, T.M. (1989), "The influence of possible fault offset on dam design", Water Power & Dam Construction,
April 1989.
Mejia, L., Forrest, M., Bischoff, J., Gillon, M., and Everitt, S. (1999), Upgrading of Matahina Dam for
foundation fault displacement, Proc. of Waterpower '99, ASCE, Las Vegas, Nevada.
Metropolitan Water District of Southern California (1996), Seismic considerations addendum report:
Characterization of two faults exposed at the West Dam, Unpubl. Rept.
Nelson, C.S. (1984), "Assessment of New Zealand Geological Survey, DSIR report on seismotectonic hazard of
the Clyde dam site", Geol. Soc. New Zealand, Newsletter, 63, pp 14-18.
Pedro, J.O. and Azevedo, M.C. (1988), Arch dam on active fault—model analysis of special construction
arrangements, Laboratorio National de Engenharia Civil, Memoria no. 722, Lisboa.
Sadler, P.M. and Rasmussen, G.S. (1991), Geologic fault evaluation, Faults A and L, Seven Oaks Dam, Santa
Ana River north of Redlands, San Bernardino County, California, Unpubl. rept. to the U.S. Army Corps of
Engineers, Los Angeles District, 8 August 1991.
Sherard, J.L., Cluff, L.S., and Allen, C.R. (1974), "Potentially active faults in dam foundations", Geotechnique,,
24, pp 367-428.
Toppozada, T.R. and Morrison, P.W. (1982), "Earthquakes and lake levels at Oroville, California", Calif. Geol.,
35, pp 115-118.
Wallace, R.E. (1984), "Eyewitness account of surface faulting during the earthquake of 28 October 1983, Borah
Peak, Idaho", Bull. Seismol. Soc. Am., 74, pp 1091-1094.
Wells, D.L. and Coppersmith, K.J. (1994), "New empirical relationships among magnitude, rupture length, and
surface displacement", Bull. Seismol. Soc. Am.,84, pp 974-1002.
Woodward-Clyde Consultants (1977), Earthquake evaluation studies in the Auburn Dam area, Rept. to the U.S.
Bureau Reclamation, 8 vols.
Woodward-Clyde Consultants (1996), Matahina Power Station seismic hazards assessment, Rept. to Electricity
Corp. of New Zealand, 2 vols.
Woodward-Clyde Consultants (1998), Matahina Dam Strengthening Project geological construction report,,
Rept. to Electricity Corp. of New Zealand, 10 July 1998.
Yeats, R. S. and Hussain, A. (1989), "Zone of late Quaternary deformation in the southern Peshawar basin,
Pakistan", Geol. Soc. Am. Spec. Paper 232, pp 265-274.
Yeats, R.S., Sieh, K., and Allen, C.R. (1997), The geology of earthquakes, Oxford Univ. Press, New York.

8 2490